{"id":"f428d82d-31a4-4278-ad7a-4f68737908f2","arxiv_id":"2607.16368","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"A facilitated Rydberg gas with an optical-pumping gain mechanism reproduces avalanche power laws, shape collapse, peak temporal correlations, and dragon-king events expected from neural criticality.","lead":"An ultracold Rydberg gas is used as a physical simulator of critical neural-network dynamics, with excitation avalanches showing power-law statistics and shape collapse near a phase transition. A controlled optical-pumping gain loop mimics metabolic resource replenishment and stabilizes the system near criticality, where temporal correlations peak and active-phase dragon-king avalanches appear.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The avalanche exponents and shape collapse — the core evidence for criticality — come from post-hoc power-law fits without uncertainties or accessible data; if a re-analysis shifts them, criteria IV–VI are not established.","rationale":"The reader's weakest assumption points to the SIS/contact-process identification, which is a real theoretical vulnerability. However, the more immediately falsifiable and load-bearing concern is the avalanche analysis that constitutes criteria IV–VI: the power-law exponents, their uncertainties, and the shape collapse are the quantitative core of the criticality demonstration. Even if the Rydberg system were a perfect SIS model, the paper's evidence would still stand or fall on whether the reported power laws survive honest statistical treatment. The absence of error bars, the post-hoc omission rules, and the unpublished dataset make this impossible to check currently. The parameter-free branching-ratio phase-boundary prediction is independent support for the presence of a phase transition, but it does not validate the avalanche exponents. The paper's own admission that the exponents are above directed-percolation values and the system is quasi-critical further weakens the claim to have 'resolved' universal criticality criteria. These considerations do not require a change from the reader's CONDITIONAL verdict: the platform claim remains plausible, but the criticality evidence is conditional on a robust re-analysis of the avalanche statistics.","tokens_in":13281,"tokens_out":7991,"duration_ms":93543,"concrete_test":"Release the raw time-resolved ion-arrival data behind Fig. 3 in a persistent repository, then re-fit P(m) and P(t) using maximum-likelihood power-law estimation with a data-driven lower cutoff, bootstrap confidence intervals, and a goodness-of-fit test against log-normal and stretched-exponential alternatives; include the gray 'insufficient statistics' points and the first two durations, or explicitly justify their exclusion. If the power-law hypothesis is rejected, or if τ_m or τ_t shift by more than ~0.1 when cuts and binning are varied, the criticality criteria (IV)–(VI) are not demonstrated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 'Proximity to a Critical Point' bases the criticality claim on Fig. 3: power-law avalanche magnitude and duration distributions, the scaling relation γ≈(τ_t−1)/(τ_m−1), and the avalanche shape collapse. The quoted exponents (τ_m=1.72, τ_t=2.31, γ=1.77) are given without uncertainties; the fits omit points labeled 'insufficient statistics' and drop the two shortest durations; and an avalanche is defined by an empty 50-μs bin. These binning and truncation choices can produce apparent power laws and a misleading collapse even in non-critical finite-size data. This is load-bearing because criteria IV–VI are the direct evidence that the system sits at a critical point. The internal consistency check γ≈(τ_t−1)/(τ_m−1) does not answer it, since all three exponents come from the same preprocessed data. The paper also acknowledges that the avalanche exponents lie 'above theoretically predicted values' and describes the system as quasi-critical, separating the observation from the SIS/directed-percolation universality class used to motivate the platform. The data-availability entry [50] is listed as unpublished, so independent re-analysis is currently impossible. If these fits are not robust, the central claim that facilitated Rydberg gases 'possess all the necessary properties to simulate critical dynamics in realistic neural networks' is not established.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental study of a Rydberg gas as a simulator for neural-network criticality. Ground-state and Rydberg atoms play the roles of inactive and active neurons, Rydberg facilitation provides the synaptic coupling, and dephasing is invoked to map the system onto a classical contact process / SIS model. The authors map two-dimensional phase diagrams, compute a mean-field branching-ratio critical line from optical Bloch equations without fitting to the activity data, and identify a critical density. At that density they report power-law avalanche magnitude and duration distributions with exponents τ_m = 1.72 and τ_t = 2.31, the scaling relation γ ≈ (τ_t − 1)/(τ_m − 1) with γ = 1.77, and a collapse of averaged avalanche shapes. They then implement an optical-pumping gain mechanism to compensate atom loss, show attraction to a steady state, measure a peak in g(2)(0) near the critical point, and observe oscillations and dragon-king avalanches slightly in the active phase. The central claim is that facilitated Rydberg gases possess all necessary properties to simulate critical dynamics in realistic neural networks.","tokens_in":13632,"tokens_out":4950,"duration_ms":63809,"significance":"If the criticality evidence is robust, this is a significant advance: it offers a tunable, time-resolved, single-particle-resolved experimental platform for studying questions from the neural-criticality literature, with a resource-replenishment channel that goes beyond most cold-atom analogues. The manuscript has real strengths: the branching-ratio calculation is an independent, non-fit anchor for the phase boundary; the phase diagrams cover a wide parameter range; and the gain mechanism is a genuinely useful control knob. The paper also connects to a defined six-criterion framework. However, the avalanche statistics and shape collapse—the direct evidence for criteria IV–VI—are presented without uncertainties and with preprocessing choices that are not shown to be innocuous, and the underlying dataset is not yet publicly available. These issues are load-bearing for the paper's strongest claims.","major_comments":[{"comment":"The avalanche exponents τ_m = 1.72, τ_t = 2.31, and γ = 1.77 are quoted without uncertainties, fit ranges, or goodness-of-fit measures. The fits omit points labeled 'insufficient statistics' and the first two durations, while an avalanche is defined by an empty 50-μs bin. These binning and truncation choices can generate apparent power laws and a misleading collapse even for non-critical finite-size data. Because criteria IV and V rest entirely on these fits, and because the text itself states that the avalanche exponents lie above theoretically predicted values and describes the system as quasi-critical, the evidence does not currently establish the claimed criticality. Please provide uncertainties, a specified fitting procedure (e.g., maximum likelihood with a stated range), robustness checks against bin width and truncation, and comparisons with alternative distributions or surrogate","section":"Proximity to a Critical Point, Fig. 3; Methods"},{"comment":"The avalanche shape collapse is presented without a quantitative measure of collapse quality. The first two durations are omitted due to temporal resolution, and the amplitude rescaling uses γ = 1.77 obtained from the same dataset. This is an internal consistency check, not an independent confirmation of universality. A quantitative collapse error, residuals, or a comparison with non-critical surrogates is needed before criterion VI can be considered resolved.","section":"Proximity to a Critical Point, Fig. 3(e)"},{"comment":"The Data Availability section states that the data are publicly available, but the cited entry [50] is currently listed as 'unpublished'. Given that the central avalanche claims depend on preprocessing choices and fits, independent re-analysis is essential. Please deposit the processed data, the raw bin counts, and the analysis code with a stable DOI, and ensure the reference is updated to a published dataset.","section":"Data Availability, Ref. [50]"},{"comment":"The branching-ratio calculation is described as having 'no free fit parameters,' but Eq. (1) depends on the calibrated values of Ω, γ, γ*, Δ, and C6, and on the assumption of a homogeneous mean-field density n. This is not a criticism of the approach—the calculation is a valuable independent anchor—but the phrase 'parameter-free' should be clarified, and a sensitivity analysis with respect to the input parameters and the mean-field assumption should be reported. Otherwise the reader cannot judge how robust the yellow critical line in Fig. 2 is.","section":"Neural Network Simulator; Methods, Eq. (1)"}],"minor_comments":[{"comment":"The fit φ = (n − n_c)^β is shown but the fitting range, background subtraction, and treatment of points far from the transition are not described. Please add these details and specify whether β is consistent with the cited directed-percolation value over the full fit range.","section":"Fig. 2(b)"},{"comment":"The caption says 'Data points with insufficient statistics (gray) have been omitted for fitting the exponent' and 'the first two durations have been omitted.' Please state the exact fit ranges and the number of points used in each fit.","section":"Fig. 3 caption"},{"comment":"The dragon-king claim rests on a visual deviation from an extrapolated power law. Please provide a statistical test (e.g., comparison of counts in the large-m interval against the fitted power-law expectation with uncertainties) and describe how the interval averaging for m > 67 is performed.","section":"Fig. 4(c)"},{"comment":"The g^(2)(0) peak is a central observation, but the inset shows points without error bars. Please add uncertainties or state that they are smaller than the symbol size.","section":"Fig. 4(b) inset"},{"comment":"The use of a single empty 50-μs bin to terminate an avalanche should be discussed explicitly in the main text, since it sets the temporal resolution of both P(t) and the shape collapse and may bias the shortest-duration bin.","section":"Methods, avalanche definition"}],"recommendation":"major_revision","confidential_remarks":"The platform and the independent branching-ratio calculation are promising, and I do not see a fundamental flaw in the experimental approach. The main obstacle is that the paper's headline criticality evidence—avalanche exponents and shape collapse—is currently presented in a way that cannot be independently checked or quantitatively assessed. With robust uncertainties, specified fits, surrogate checks, and a publicly available dataset, the claims could become publishable. I would not reject at this stage, but the revision must address the statistical robustness of criteria IV–VI and the data-availability gap."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new thing here is the controlled gain channel: optical pumping from a reservoir state to replace lost atoms, which lets them hold the system in a steady state near the absorbing transition and study long-time dynamics. That is a real experimental step beyond earlier Rydberg SOC work. The phase diagram is carefully mapped, the independent branching-ratio calculation is a credible anchor for the transition line, and the g(2) peak near criticality is a nice, clean observable. I believe the platform claim — a tunable atomic simulator for resource-replenished critical dynamics — is largely established.\n\nThe soft spots are where the stress-test note lands, and they are real. The avalanche exponents τ_m = 1.72, τ_t = 2.31, γ = 1.77 are quoted without uncertainties, the fits drop short durations and points labeled \"insufficient statistics,\" and the avalanche definition depends on a 50 μs bin. Those choices can manufacture apparent power laws and a shape collapse even away from criticality. The internal scaling check γ ≈ (τ_t−1)/(τ_m−1) does not rescue this, because all three exponents come from the same preprocessed data. The dragon-king claim in Fig. 4(c) is just visual — no statistical test, no comparison to a null model. And the data availability statement is not true in practice: reference [50] is listed as unpublished, so no one can re-analyze the fits. That is a concrete, fixable problem.\n\nTo the paper's credit, the authors acknowledge the avalanche exponents sit above directed percolation values and use the word \"quasi-critical.\" That is honest. But it undercuts the abstract's \"resolve criticality criteria\" and the claim that the system possesses \"all necessary properties\" for neural criticality. The SIS mapping via dephasing is plausible but not rigorously checked; if dephasing is incomplete or blockade effects creep in, the observed statistics may not be universal.\n\nBottom line: this is a capable experimental group with a good platform and a promising control knob. The paper deserves peer review, but the referees should push for uncertainty estimates on the exponents, a robustness check of the fits to binning and truncation, a statistical test for the dragon-king excess, and actual public data. If those come, the paper would be much stronger. As it stands, I would cite it for the gain mechanism and phase diagram, but not for the avalanche exponents.","headline":"A solid Rydberg platform paper whose central criticality evidence hinges on avalanche fits that lack uncertainties and accessible data — worth refereeing, but the neural-network claims need to be softened or the fits hardened.","tokens_in":14110,"tokens_out":1161,"would_cite":true,"duration_ms":15156,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims that a facilitated ultracold Rydberg gas can serve as a controlled physical simulator of neural-network criticality, reproducing power-law avalanches, universal shape collapse, and resource-driven oscillations near a non-e","keywords":["neural criticality","Rydberg facilitation","ultracold gases","non-equilibrium phase transitions","avalanche criticality","self-organized criticality","dragon king avalanches","contact process"],"falsifier":"Measure the avalanche distributions at substantially finer time binning (e.g., 10 microseconds rather than 50 microseconds) and verify that the exponents tau_m, tau_t, and the shape collapse are independent of bin width; or directly measure the dephasing rate through coherence spectroscopy and check whether it is large enough to justify the classical contact-process mapping. If exponents shift systematically with binning, or if the branching ratio computed from independently measured two-atom facilitation probabilities deviates from 1 at the apparent critical density, the central claim would b","tokens_in":13185,"feed_emoji":"🧠","tokens_out":5153,"duration_ms":58446,"temperature":0.7,"pith_summary":"The paper tries to establish that an ultracold gas of rubidium atoms coupled by Rydberg facilitation is a faithful, controllable simulator of the critical dynamics believed to underlie efficient neural networks. The central move is to identify each atom with a neuron, a facilitated excitation with a synaptic spike, atom loss with metabolic resource consumption, and optical pumping with resource replenishment. Against this mapping, the experiment measures all six criteria for criticality in one platform: an absorbing-to-active phase transition with branching ratio one, power-law avalanche statistics obeying the scaling relation, and collapse of avalanche shapes onto one universal curve. It also finds that correlations peak at the critical point and that, slightly in the active phase, the system shows stochastic oscillations and dragon-king avalanches, matching predictions for systems orbiting criticality. If correct, this gives experimental access to questions about cortical criticality that biological data alone cannot answer because of undersampling and lack of control.","feed_headline":"Rydberg atoms simulate neural-network criticality","feed_subtitle":"A tunable gas of ultracold rubidium shows avalanche power laws, shape collapse, and gain-driven oscillations.","key_machinery":"Rydberg facilitation, the distance-selective resonance at r_fac = (C6/Delta)^(1/6) at which an excited atom switches a nearby ground-state atom into resonance, creates dynamic local synaptic connections; strong dephasing renders the resulting spreading a classical contact process, mapping the gas to the susceptible-infected-susceptible epidemic model with an absorbing-state phase transition. An optical-pumping gain from a reservoir hyperfine state replenishes lost atoms and tunes the steady state, while a parameter-free mean-field branching-ratio integral over the optical Bloch equations predicts the critical density where the branching ratio equals one.","core_discovery":"The experiments demonstrate that a driven, dissipative Rydberg gas with a controlled gain channel exhibits a non-equilibrium phase transition of the same type as the neural SIS/contact-process class. Tuning the ground-state density across a predicted critical density (branching ratio = 1, derived from a parameter-free mean-field integral over the optical Bloch equations) yields a power-law order parameter with exponent beta near 0.78, avalanche magnitude and duration distributions consistent with power laws (tau_m near 1.72, tau_t near 2.31), a magnitude-duration exponent gamma near 1.77 obeying the scaling relation, and collapse of rescaled avalanche shapes. With optical pumping acting as a","pith_inferences":["Beyond the paper: if the dephasing rate can be dialed down, the same platform should cross over from the classical contact-process regime to coherent quantum dynamics; a clean observation of that crossover would test the paper's load-bearing dephasing assumption and open a route to quantum neural networks.","Beyond the paper: because the gain rate is a continuous knob, one could search directly for self-organized criticality by tuning gain to match loss and watching whether the system tunes itself to the critical density without external parameter adjustment.","Beyond the paper: a spatially resolved version of the gain channel could emulate glial resource transport and test whether spatially inhomogeneous replenishment shifts or destroys the critical point, a question the current global-pump setup leaves open."],"forward_implications":["If the mapping holds, a tabletop atomic system becomes a testbed in which the neural criticality hypothesis can be examined under repeatable, tunable conditions that biological experiments cannot provide.","Density, laser detuning, and Rabi frequency act as independent control knobs for the microscopic spreading process, so the phase transition can be crossed in several ways and compared to parameter-free mean-field predictions.","The optical-pumping gain mimics metabolic resource replenishment, implying that resource dynamics alone can stabilize quasi-critical operation and generate long-time oscillations in an excitable medium.","Peak g(2)(0) near the transition offers a correlation-based estimator of criticality that could be applied to spiking data from real neural networks.","Reproduction of dragon-king avalanches in a physical system suggests these extreme events are generic features of resource-limited systems orbiting criticality, not peculiar to biological wiring."],"fun_headline_variants":["Rydberg gas simulates neural criticality","Avalanche scaling in Rydberg gas mimics brain dynamics","Gain-stabilized Rydberg gas shows critical avalanches","Ultracold atoms replicate neural network phase transitions"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"The interpretation of the observed avalanches as critical phenomena rests on strong dephasing making excitation spreading a classical Markovian contact process; if dephasing is incomplete or long-range interactions and binning distort the dynamics, the power laws and shape collapse lose their stated criticality interpretation.","fun_headline_variants_meta":{"raw":{"variants":["Rydberg gas simulates neural criticality","Avalanche scaling in Rydberg gas mimics brain dynamics","Gain-stabilized Rydberg gas shows critical avalanches","Ultracold atoms replicate neural network phase transitions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000159,"raw_usage":{"total_tokens":1051,"prompt_tokens":717,"completion_tokens":334,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":461,"completion_tokens_details":{"reasoning_tokens":268}},"tokens_in":461,"tokens_out":334,"duration_ms":4359,"temperature":1.0,"reasoning_tokens":268,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T21:44:40.560934+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the avalanche distributions at substantially finer time binning (e.g., 10 microseconds rather than 50 microseconds) and verify that the exponents tau_m, tau_t, and the shape collapse are independent of bin width; or directly measure the dephasing rate through coherence spectroscopy and check whether it is large enough to justify the classical contact-process mapping. If exponents shift systematically with binning, or if the branching ratio computed from independently measured two-atom facilitation probabilities deviates from 1 at the apparent critical density, the central claim would b","supporting_citations":[],"review_version":1}